KR20120024686A - Abnormality detection system, abnormality detection method, and recording medium - Google Patents

Abstract

An abnormality detection system for accurately detecting abnormality occurring in an apparatus is provided. The abnormality detection system 100 which detects the abnormality which arises in the plasma processing apparatus 2 has each of the several ultrasonic sensors 41 which detect AE resulting from abnormality, and each output signal of the ultrasonic sensor 41, respectively. A divider 65 for distributing the first signal and the second signal, a trigger 52 for generating a trigger signal when the first signal is sampled at 10 kHz, for example, and detecting a predetermined characteristic, and receiving and triggering the trigger signal. A trigger determined from a trigger generation time counter 54 for determining an occurrence time, a data logger board 55 for creating sampling data obtained by sampling the second signal at, for example, 1 MHz, and a trigger occurrence time counter 54 among sampling data. The PC 50 which analyzes the abnormality which arose in the plasma processing apparatus 2 by performing waveform analysis of the data corresponded to a fixed period based on the generation time is provided.

Description

ABNORMALITY DETECTION SYSTEM, ABNORMALITY DETECTION METHOD, AND RECORDING MEDIUM

The present invention relates to an abnormality detection system, an abnormality detection method by an abnormality detection system, and a computer-readable storage medium storing a program used in the abnormality detection system.

BACKGROUND OF THE INVENTION A plasma processing apparatus that performs a predetermined plasma processing on a substrate, such as a semiconductor wafer or a flat panel display panel, usually includes a storage chamber (hereinafter referred to as "chamber") for accommodating a substrate. In such a substrate processing apparatus, plasma is generated from the processing gas by applying high frequency power into the chamber while introducing the processing gas into the chamber, and the plasma is subjected to the plasma processing by the plasma.

When high frequency power is applied to the chamber, abnormalities such as plasma abnormal discharge (eg, micro arcing) may occur due to various factors. Plasma abnormal discharge causes cracks, notches, or the like on the surface of the substrate, or burns out component parts disposed in the chamber, and in addition, component parts (for example, upper electrodes) in the chamber. It becomes the cause of particle | grains by peeling off deposits etc. which adhered to it.

Therefore, when plasma abnormal discharge is detected early, and when plasma abnormal discharge is detected, it is necessary to promptly take appropriate measures such as stopping the operation of the plasma processing apparatus to prevent damage to the substrate or component parts and generation of particles. There is. Therefore, various methods for early detection of abnormalities such as plasma abnormal discharges have been developed.

For example, as a method capable of high sensitivity detection of the plasma processing method, a method of detecting AE (Acoustic Emission) due to energy release during plasma abnormal discharge has been studied. A detection device using AE, comprising a plurality of ultrasonic sensors on the outer wall of the chamber, for detecting AE resulting from energy release when a plasma abnormal discharge occurs by these ultrasonic sensors, or a mounting table for loading a semiconductor wafer. A device is known in which a plurality of acoustic probes are provided so as to contact a focus ring arranged around a semiconductor wafer loaded on a susceptor or a mounting table, and the ultrasonic waves propagating these acoustic probes are detected by an ultrasonic sensor (for example, , Patent Document 1). At this time, the method of monitoring the high frequency power (voltage or current) used for plasma generation may be used together.

Patent Document 1: Japanese Patent Publication No. 2003-100714

However, the ultrasonic sensor detects not only AE caused by plasma abnormal discharge but also mechanical vibration and the like caused by opening and closing of the gate valve of the plasma processing apparatus as noise. This causes a problem that the detection accuracy of the plasma abnormal discharge is lowered. In order to solve this problem, improvement of the analysis method of the AE signal detected by the ultrasonic sensor is required.

As an analysis method of an AE signal, for example, there is a method of digitally processing data obtained by high-speed sampling of an output signal (detection signal) of an ultrasonic sensor on a personal computer (PC). However, in this method, since it is necessary to handle large-capacity data by high-speed sampling, there are problems such as high data processing costs and inability to perform real-time data processing.

There is also a method of performing digital processing in combination with a mock-up. However, due to limitations in the processing capability of a DSP (Digital Signal Processor), it corresponds to 10 kHz sampling. This sampling frequency is sufficient for monitoring high frequency power, but is insufficient for monitoring the AE signal due to an abnormality of about microseconds from the time of generation to the end.

By the way, conventionally, even when the some ultrasonic sensor is arrange | positioned, analysis of AE signal is performed for every ultrasonic sensor. In this case, for example, since a difference occurs in the magnitude of the AE signal due to the difference in the generation site of the plasma abnormal discharge, there is a possibility that the occurrence of the plasma abnormal discharge may be missed.

An object of the present invention is an abnormality detection system and abnormality detection system capable of detecting data abnormality with high accuracy while acquiring a large amount of data relating to an abnormality occurring in a processing device, and detecting a generated abnormality with high accuracy. The present invention provides a computer-readable storage medium storing a method and a program used in an abnormality detection system.

In order to achieve the above object, the abnormality detection system according to the present invention is an abnormality detection system for detecting abnormality occurring in the processing apparatus, comprising: a plurality of ultrasonic sensors for detecting acoustic emission generated in the processing apparatus, and A distribution unit for distributing each output signal of the plurality of ultrasonic sensors into a first signal and a second signal, respectively, and a trigger generation unit for generating a trigger signal when sampling the first signal at a first frequency and detecting a predetermined characteristic; And a trigger generation time determination unit that receives the trigger signal to determine a trigger generation time, a data creation unit that creates sampling data obtained by sampling the second signal at a second frequency higher than the first frequency, and the sampling. Based on the trigger occurrence time determined by the trigger occurrence time determination unit among data By performing a waveform analysis of the data corresponding to a period of time as characterized by a data processing unit to interpret the above occurred in the processing device.

The abnormality detection system further includes a trigger signal processing unit for integrating the plurality of trigger signals into one signal when a plurality of the trigger signals are generated within a predetermined period, and determining the trigger generation time. The unit may determine the trigger generation time with respect to the representative trigger signal.

The abnormality detection system may further include a filter for removing noise from each output signal of the plurality of ultrasonic sensors.

The abnormality detection system is characterized in that the first frequency is 10 kHz to 5 MHz, and the second frequency is 500 kHz to 5 MHz.

In order to achieve the above object, the abnormality detecting method according to the present invention is an abnormality detecting method for detecting abnormality occurring in the processing apparatus, comprising: a detecting step of detecting acoustic emission generated in the processing apparatus by a plurality of ultrasonic sensors; A distribution step of distributing each output signal from the plurality of ultrasonic sensors obtained in the detection step by a distribution unit into a first signal and a second signal, respectively, and an A / D conversion unit at the first frequency A trigger signal generation step of generating a trigger signal by the signal generating unit when sampling by a predetermined characteristic is detected, and a trigger generation that receives the trigger signal and determines the trigger generation time of the trigger signal by the time counter unit. And determining the time, and converting the second signal to a second frequency higher than the first frequency. The computer performs a sampling data creation step of sampling by a net and creating sampling data, and a waveform analysis of data corresponding to a predetermined period of time based on the trigger generation time determined in the trigger generation time determination step of the sampling data. And a data processing step of analyzing an abnormality occurring in the processing apparatus.

Further, the abnormality detection method according to the present invention, when a plurality of the trigger signal is generated within a predetermined period in the trigger signal generation step, the trigger signal processing to integrate the plurality of trigger signals into a single signal as a representative trigger signal The method further includes the step of determining the trigger generation time, wherein the trigger generation time is determined with respect to the representative trigger signal.

In addition, the abnormality detection method according to the present invention is characterized in that it further comprises a noise removing step of removing noise by a filter from the first signal and the second signal obtained by the distribution step.

The abnormality detection method according to the present invention is characterized in that the first frequency is 10 kHz to 5 MHz, and the second frequency is 500 kHz to 5 MHz.

In the abnormality detection method according to the present invention, the data processing step includes a cutting step of cutting out data corresponding to the predetermined period from the sampling data, and a data cut down by a representative value with respect to the data cut out in the cutting step. In the case where a meaningful waveform exists in the generated down sampling data by sampling, a first waveform feature variable extracting step of extracting a waveform feature variable from the down sampling data and the first waveform feature variable extracting step are extracted. A second waveform feature variable extraction step of narrowing an analysis target of the data cut out in the cutting step by estimating the time of the waveform feature variable, and extracting a waveform feature quantity from the data cut out in the cutting step with respect to the analysis target; Waveform Feature Amount and Preset Abnormality Obtained in the Second Waveform Feature Amount Extraction Step And a determination step of determining an abnormality occurring in the processing apparatus by performing pattern recognition with a pattern recognition model.

Moreover, the abnormality detection method which concerns on this invention is a abnormality detection method as described in any one of Claims 5-9, The process condition which acquires the process conditions of the predetermined process performed in the said processing apparatus made before the said detection step. The method may further include an acquiring step, wherein the detecting step is executed only during the execution period of the predetermined process included in the process condition acquired in the process condition acquiring step.

In order to achieve the above object, a computer readable storage medium according to the present invention is a computer readable medium for storing a program for executing an abnormal detection method for detecting an abnormality occurring in a processing apparatus determined by a computer-controlled abnormality detection system. As a possible storage medium, the abnormality detection method includes a detection step of detecting acoustic emission generated by the processing apparatus by a plurality of ultrasonic sensors and a respective detection signal from the plurality of ultrasonic sensors obtained in the detection step. A distribution step of distributing the first signal and the second signal by the distribution unit, and sampling the first signal at a first frequency by the A / D conversion unit to detect a predetermined feature and then triggering the trigger signal by the signal generation unit. Generating a trigger signal and receiving the trigger signal to generate the trigger signal. A trigger generation time determination step of determining a trigger generation time of a trigger signal by a time counter unit, and sampling the second signal at a second frequency higher than the first frequency by the A / D conversion unit to create sampling data; Analyzing the abnormality that has occurred in the processing apparatus by performing a waveform analysis of data corresponding to a predetermined period of time based on the trigger generation time determined in the trigger generation time determination step of the sampling data and the sampling data; A data processing step.

According to the abnormality detecting system according to the present invention, the abnormality detecting method according to claim 5, and the computer-readable storage medium according to claim 11, the data processing cost can be kept low while acquiring a large amount of data relating to the abnormality occurring in the processing apparatus. Moreover, the abnormality which occurred can be detected with high precision.

In addition, according to the abnormality detection system and the abnormality detection method according to claim 6 of the present invention, by generating a representative trigger signal, the load on the data processing can be reduced without lowering the accuracy of the abnormality detection.

Moreover, according to the abnormality detection method which concerns on this invention, since a noise is removed by the filter from each output signal of an ultrasonic sensor, the generation of a trigger signal and the analysis precision of sampling data can be improved.

Moreover, according to the abnormality detection system and the abnormality detection method which concern on this invention, it becomes possible to detect the AE signal resulting from the abnormality which was difficult to detect conventionally of generation | occurrence | production to end in microseconds after generation | occurrence | production.

In addition, according to the abnormality detection method which concerns on this invention, since the data amount which becomes an analysis process object is reduced by data truncation, since down-sampled data with a small data amount is used and the object of an analysis process is narrowed down, abnormality detection is carried out. The data processing time can be shortened without lowering the accuracy of the.

Further, according to the abnormality detection method according to the present invention, since data is acquired in the detection step only during the execution period of the predetermined process included in the process condition, the load of the overall data processing can be reduced, and the detected Analysis and determination which narrowed the above kind can be performed.

BRIEF DESCRIPTION OF THE DRAWINGS It is sectional drawing which shows schematic structure of the plasma processing apparatus to which the abnormality detection system which concerns on this invention is applied.
2 is a cross-sectional view illustrating a schematic configuration of an ultrasonic sensor.
It is a schematic block diagram of the abnormality detection system which concerns on embodiment of this invention.
It is a flowchart which shows the outline | summary of the operation form of the abnormality detection system.
5 is a flowchart showing a schematic procedure of an abnormality detection / analysis processing method by the abnormality detection system.
FIG. 6 is a flowchart showing the detailed procedure of step S23 (trigger output) in FIG.
FIG. 7 is a flowchart showing the detailed procedure of step S24 (waveform cut / preserve) in FIG. 5.
FIG. 8 is a flowchart showing the detailed procedure of step S25 (waveform analysis) in FIG.
FIG. 9 is a flowchart showing the detailed procedure of step S27 (abnormal pattern determination) in FIG.

EMBODIMENT OF THE INVENTION Hereinafter, embodiment of this invention is described, referring drawings.

BRIEF DESCRIPTION OF THE DRAWINGS It is sectional drawing which shows schematic structure of the plasma processing apparatus to which the abnormality detection system which concerns on embodiment of this invention is applied. The plasma processing apparatus 2 is subjected to an etching process on a semiconductor wafer (hereinafter referred to as "wafer") W, and includes a cylindrical chamber 10 made of metal such as aluminum or stainless steel. In the chamber 10, for example, a cylindrical susceptor 11 as a stage for loading a wafer W having a diameter of 300 mm is provided.

The chamber 10 is provided with the maintenance opening part (not shown) which communicates the inside and exterior of the chamber 10, and the maintenance cover part (not shown) which opens and closes this opening part freely. In addition, between the side wall of the chamber 10 and the susceptor 11, an exhaust path 12 serving as a flow path for discharging gas above the susceptor 11 out of the chamber 10 is provided. An annular exhaust plate 13 is arranged in the middle of the exhaust passage 12, and the space downstream of the exhaust plate 13 is an APC (Adaptive Pressure Control Valve) 14, which is a variable butterfly valve. Through. The APC 14 is connected to a TMP (Turbo Molecular Pump) 15, which is an exhaust pump for vacuum operation, and the TMP 15 is connected to a DP (Dry Pump: Dry Pump) 16, which is an exhaust pump. have.

In addition, the exhaust flow path comprised by the APC 14, TMP 15, and DP 16 is called "main exhaust line" hereafter. In this main exhaust line, the pressure control in the chamber 10 is performed by the APC 14, and the pressure inside the chamber 10 can be reduced in a high vacuum state by the TMP 15 and the DP 16. FIG.

The space downstream from the exhaust plate 13 is communicated with the DP 16 via an exhaust flow path (hereinafter referred to as a "roughing line") separate from the main exhaust line. The roughing line has, for example, an exhaust pipe 17 having a diameter of 25 mm and a valve V2 arranged in the middle of the exhaust pipe 17. When the DP 16 is driven, the roughing line passes through the roughing line, and the chamber ( 10) It is possible to exhaust the gas inside.

The susceptor 11 is connected to a high frequency power source 18 for supplying a predetermined high frequency power to the susceptor 11 through a feed rod 40 and a matching device 19. Accordingly, the susceptor 11 functions as a lower electrode. In addition, the matching unit 19 reduces the reflection of the high frequency power from the susceptor 11 and increases the supply efficiency of the high frequency power to the susceptor 11. On the other hand, the power output from the high frequency power source 18 is monitored by a current sensor or a voltage sensor (not shown).

A disk-shaped electrode plate 20 including a conductive film is disposed above the inside of the susceptor 11 so as to adsorb the wafer W with an electrostatic attraction force, and the DC plate 22 is provided on the electrode plate 20. This is electrically connected. The wafer W is adsorbed and held on the upper surface of the susceptor 11 by a Coulomb force or a Johnson-Rahbek force generated by a DC voltage applied from the DC power supply 22 to the electrode plate 20. do. Further, above the susceptor 11, a circle containing silicon (Si) or the like is used in order to converge the plasma generated in the space S above the susceptor 11 toward the wafer W. An annular focus ring 24 is arranged.

Inside the susceptor 11, a coolant chamber 25 is provided. Here, the coolant chamber 25 is formed in an annular shape extending in the circumferential direction. In the refrigerant chamber 25, a refrigerant (eg, cooling water) having a predetermined temperature is circulated and supplied from a chiller unit (not shown) through a pipe 26, and the wafer is disposed on the susceptor 11 by the temperature of the refrigerant. The processing temperature of (W) is controlled.

In the upper surface of the susceptor 11, a portion of the wafer W adsorbed (hereinafter referred to as an "adsorption surface") includes a plurality of heat transfer gas supply holes 27 and a heat transfer gas supply groove (not shown). Formed. The heat transfer gas supply hole 27 and the like are connected to the heat transfer gas supply unit 29 through the heat transfer gas supply line 28 provided in the susceptor 11, and the heat transfer gas supply unit 29 is a heat transfer gas such as He gas. Is supplied to the gap between the adsorption surface and the back surface of the wafer (W). On the other hand, the electrothermal gas supply part 29 is connected to the DP 16 so that the clearance gap between a suction surface and the back surface of the wafer W may be made into a vacuum state.

On the suction surface of the susceptor 11, a plurality of pusher pins 30 are provided as lift pins which protrude freely from the upper surface of the susceptor 11. The pusher pin 30 is movable in the vertical direction of FIG. 1 by converting the rotational motion of a motor (not shown) into linear motion by a ball screw or the like. When the wafer W is adsorbed and held on the suction surface, the pusher pin 30 is accommodated in the susceptor 11, and when the wafer W is brought in or taken out of the chamber 10, the pusher pin 30 is It protrudes from the upper surface of the susceptor 11, and lifts the wafer W upwards apart from the susceptor 11.

The shower head 33 is disposed at the ceiling of the chamber 10. The high frequency power source 21 is connected to the shower head 33 via the matching unit 23, and the high frequency power source 21 supplies the high frequency power to the shower head 33. Thus, the shower head 33 functions as an upper electrode. On the other hand, the function of the matching device 23 is the same as that of the matching device 19 described above.

The power output from the high frequency power supply 21 is monitored by a current sensor or a voltage sensor (not shown).

The shower head 33 is disposed on the lower surface side thereof, and includes an electrode plate 35 having a plurality of gas vent holes 34, and an electrode support 36 that detachably supports the electrode plate 35. A buffer chamber 37 is formed inside the electrode support 36, and the buffer chamber 37 and a processing gas supply unit (not shown) are connected by a processing gas introduction pipe (piping) 38. . A pipe insulator 39 is disposed in the middle of the process gas introduction pipe 38, the pipe insulator 39 includes an insulator, and the high frequency electric power supplied to the shower head 33 passes through the process gas introduction pipe 38. Prevents flow to the process gas supply.

The gate valve 5 which opens and closes the carrying-in / out port 31 of the wafer W is attached to the side wall of the chamber 10.

In the plasma processing apparatus 2, the high frequency power is supplied to the susceptor 11 and the shower head 33, respectively, and the shower head 33 is provided in the space S between the susceptor 11 and the shower head 33. By supplying the processing gas from the chamber, a high-density plasma containing ions and radicals is generated in the space S.

In the plasma processing apparatus 2, when the etching process is performed, first, the gate valve 5 is opened, and the wafer W to be processed is loaded into the chamber 10 and loaded on the susceptor 11. Subsequently, a DC voltage is applied from the DC power supply 22 to the electrode plate 20 to adsorb the wafer W onto the susceptor 11.

Thereafter, a process gas (eg, a mixed gas including C ₂ F ₈ gas, O ₂ gas, and Ar gas at a predetermined flow rate ratio) is introduced into the chamber 10 from the shower head 33 at a predetermined flow rate and flow rate ratio, The pressure in the chamber 10 is set to a predetermined value by the main exhaust line. In addition, high frequency power is applied to the chamber 10 by the susceptor 11 and the shower head 33. In this way, in the space S, the processing gas is converted into plasma, and the generated radicals and ions are condensed on the surface of the wafer W by the focus ring 24, so that the surface of the wafer W is physical or Chemically etched.

At this time, when a plasma abnormal discharge such as micro arcing occurs, it is detected by using an ultrasonic sensor to detect AE resulting from the energy release caused by the plasma abnormal discharge. An ultrasonic sensor is one of the components of the abnormality detection system 100 mentioned later.

2 is a cross-sectional view illustrating a schematic configuration of an ultrasonic sensor. The ultrasonic sensor 41 is mounted on a water wave plate 42 of a plate type including an insulator such as Al ₂ O ₃ , and a metal film 43 such as a silver deposition film on the water wave plate 42. A piezoelectric element (e.g., lead zirconate titanate-based piezoelectric ceramic) 44 and a case type including a metal (e.g., aluminum or stainless steel) mounted on the water wave plate 42 to cover the piezo element 44. The shield case 45 is provided.

When the piezoelectric element 44 receives a physical vibration such as an ultrasonic wave, the piezoelectric element 44 generates a voltage having a magnitude corresponding to the magnitude of the vibration. In order to extract this voltage signal, a connector 46 exposed to the inside and the outside of the shield case 45 is disposed on the side wall of the shield case 45, so that the metal film 47 and the connector 46 have internal wiring ( 48, an external wiring 49 is connected to the connector 46, and a voltage signal generated by the piezoelectric element 44 is taken out through the external wiring 49.

The ultrasonic sensor 41 is mounted outside the component, for example, the chamber 10 or the pipe insulator 39, in which the occurrence of plasma abnormal discharge is predicted in the plasma processing apparatus 2. Specifically, in order to detect the ultrasonic wave propagated to the outer wall of the chamber 10 due to the occurrence of the plasma abnormal discharge, the ultrasonic wave sensor 42 is brought into close contact with the outer wall of the chamber 10. To the chamber 10.

On the other hand, depending on the component of the plasma processing apparatus 2, a leakage current flows from the component to the ultrasonic sensor 41, which may prevent the ultrasonic sensor 41 from accurately detecting abnormal discharge. However, in the ultrasonic sensor 41, this problem can be avoided because the water wave plate 42 including the insulator blocks the leakage current. The insulator used for the water wave plate 42 may transmit ultrasonic waves, and is not limited to Al ₂ O ₃ .

Next, an abnormality detection system such as plasma abnormal discharge generated in the plasma processing apparatus 2 will be described. It is a schematic block diagram of the abnormality detection system which concerns on embodiment of this invention.

In the plasma processing apparatus 2, since the power output from the high frequency power sources 18 and 21 is monitored as described above, this monitor signal can be used for abnormality detection (mainly abnormality detection during plasma generation). However, as described below, the sampling frequency of the monitor signal is set to 10 kHz, for example, whereas the sampling frequency of the output signal of the ultrasonic sensor 41 (hereinafter referred to as the "sensor signal") is set to 1 MHz, for example. The sensor signal is obtained in detail. Therefore, in this case, the abnormality generated in the plasma processing apparatus 2 is detected and analyzed based on the sensor signal (hereinafter referred to as "abnormal detection / analysis processing"), and the monitor signal is assisted in abnormal detection / analysis. It shall be used.

The abnormality detection system 100 includes an ultrasonic sensor 41 disposed in the plasma processing apparatus 2, a distributor 65 for distributing sensor signals from the ultrasonic sensor 41 into two identical signals, and a distributor ( Filter 51a for removing noise from one sensor signal (first signal) output from 65 and for removing noise from the other sensor signal (second signal) output from divider 65. The trigger 51 which detects a predetermined characteristic included in the filter 51b, the sensor signal passing through the filter 51a, and generates a trigger signal, and the trigger signal output from the trigger 52 are determined. An OR circuit 53 for performing an operation (processing) and a PC (Personal Computer: Personal Computer) 50 for data processing the output signal from the OR circuit 53 and the sensor signal passing through the filter 51b are provided. .

The abnormality detection system 100 is equipped with the some ultrasonic sensor 41, Therefore, even if the abnormality which is difficult to find by the information from one ultrasonic sensor 41, the other ultrasonic sensor 41 differs in the abnormality. It is possible to increase the probability of strong detection. That is, high detection accuracy is realized by performing multivariate (multichannel) analysis for analyzing information from the plurality of ultrasonic sensors 41. Although the number of the ultrasonic sensors 41 is four here, it is not limited to this.

In the abnormality detection system 100, the sensor signal output from the ultrasonic sensor 41 is distributed by the distributor 65 into the same signal of two systems. This is for the purpose of performing data processing of a sensor signal efficiently as mentioned later.

The sensor signal output from the ultrasonic sensor 41 is an analog voltage signal. This sensor signal includes mechanical vibrations and the like generated by the operation of the plasma processing apparatus 2 as noise, and in many cases, the wavelength is different from AE showing such noise and abnormality such as plasma abnormal discharge to be detected. Thus, one sensor signal output from the distributor 65 is passed through a filter 51a (specifically, a high pass filter (HPF)) so as to block unnecessary low frequency noise before input to the trigger 52. The filter 51a contributes to the determination of the trigger output condition in the trigger 52.

On the other hand, before the other sensor signal output from the distributor 65 is input to the data logger board 55 (described later) of the PC 50, the filter 51b (to block unnecessary low frequency noise) ( Specifically, it passes through a BPF (Band Pass Filter). The filter 51b contributes to the calculation of the characteristic amount of the frequency relationship which becomes the abnormality determination material in the abnormality detection / analysis process performed in the PC 50.

The trigger 52 is hardware (H / W) that judges whether or not the ultrasonic sensor 41 has detected an abnormality by simply analyzing a sensor signal and extracting a predetermined feature included in the sensor signal. On the other hand, the trigger 52 is provided with software and condition input means (for example, an operation panel) for changing and setting the operating conditions of H / W.

The trigger 52 has a function as an A / D converter for sampling each sensor signal at a sampling frequency of 10 kHz, for example. The sampling frequency at the trigger 52 can be selected in the range of 10 kHz to 5 MHz.

The trigger 52 has a function as a signal generator, and when it is determined that no abnormality has occurred in the plasma processing apparatus 2, a signal of "0" indicating that the abnormality has a predetermined characteristic suspected of abnormality has occurred. If found, a signal of " 1 " For example, the trigger 52 determines that an abnormality has occurred in the plasma processing apparatus 2 when a peak showing a maximum value larger than a preset threshold is detected, and generates a trigger signal.

The output lines of the trigger 52 may be provided for each sensor signal or may be integrated into one. Here, the former configuration is assumed. The OR circuit 53 is provided with four system input lines corresponding to the output lines of the trigger 52 provided for every four ultrasonic sensors 41.

The OR circuit 53 is an H / W that inspects a signal sent from the trigger 52 and performs a process by the first or second trigger signal processing method described below when the trigger signal is received.

The first trigger signal processing method by the OR circuit 53, upon receiving the trigger signal from the trigger 52, irrespective of which input line the trigger signal is, all the trigger signals are delayed by the PC substantially without delay along the time series. It is a method of transmitting to (50). In the first method, even if a large number of trigger signals occur, such as a case where the number of ultrasonic sensors 41 is small (for example, two), the increase in the load of the abnormality detection / analysis processing in the PC 50 is limited. It is preferable to use in case of.

The second trigger signal processing method by the OR circuit 53 integrates the trigger signal generated within a certain period into one trigger signal to the PC 50 regardless of which of the four input lines has been received. How to print

Specifically, when the OR circuit 53 receives the first trigger signal from the trigger 52, the first trigger signals all the trigger signals received within a predetermined period (hereinafter referred to as an "integrated period") from the reception time. It is assumed that the cause is the same as the abnormality that caused the signal, and a trigger signal (hereinafter referred to as the "representative trigger signal") in which these trigger signals are integrated is generated, and the representative trigger signal is transmitted to the PC 50. Output The trigger signal that is first received after the representative trigger signal is output to the PC 50 becomes the start reference of the next integration period.

In either of the first and second trigger signal processing methods, since the output line from the OR circuit 53 to the PC 50 may be one line, the OR circuit 53 is provided with one line output line. In addition, you may comprise the trigger 52 and OR circuit 53 as one H / W.

The PC 50 detects an abnormality for specifying an abnormality generated in the plasma processing apparatus 2 based on a trigger signal or a representative trigger signal output from the OR circuit 53 and a sensor signal passed through the filter 51b. / Analysis process is performed.

Hereinafter, in the abnormality detection system 100, the OR circuit 53 outputs the representative trigger signal according to the above-mentioned 2nd trigger signal processing method, and demonstrates the structure of a PC 50, a data processing method, etc.

The PC 50 includes a CPU 59, a RAM 58 for temporarily storing program data and data to be arithmetic in order to perform abnormality detection / analysis processing, a boot program or an operating system (OS). ROM 57 for storing a program and the like, and a hard disk drive (HDD) 61 which is a storage device for storing programs and data used for abnormality detection / analysis processing, intermediate data obtained during abnormality analysis, analysis results, and the like. Equipped with.

In addition to these, the PC 50 may include, for example, an input means such as a keyboard or a mouse, a liquid crystal display (LCD) as a monitor, a graphics board, a drive that handles a storage medium such as a CD-ROM or a DVD-RAM, a LAN or the Internet. And an interface for connecting to a communication line.

In addition, the PC 50 receives the representative trigger signal output from the OR circuit 53, and triggers a time counter 54 for specifying the time at which the representative trigger signal was generated (hereinafter referred to as "trigger generation time"). And a data logger board 55 for digitally sampling the four sensor signals passing through the filter 51b at a predetermined frequency for each sensor signal and storing them as digital data, and a high frequency power source 19, 21. And a data logger board 56 for digitally sampling the monitor signal at a predetermined frequency and storing it as digital data.

The trigger occurrence time counter 54 and the data logger boards 55 and 56 are each mounted on, for example, a PCI bus of the PC 50, and operation control is performed by a driver installed in the PC 50.

The trigger generation time counter 54 includes an internal clock, and recognizes, for example, the time at which the representative trigger signal is received as the trigger generation time. As described above, the representative trigger signal is generated when the integration period has elapsed from the time of generation of the first trigger signal from the trigger 52 serving as the reference for generating the representative trigger signal. Therefore, a time lug occurs between the time when abnormality actually occurred in the plasma processing apparatus 2, and a trigger generation time. Therefore, as will be described later, the PC 50 performs an abnormality detection / analysis process in consideration of this time lug.

The data logger board 55 has a function as an A / D converter. The data logger board 55 performs high-speed sampling of a sensor signal passing through the filter 51b at a frequency of, for example, 1 MHz, for example, digital data (hereinafter referred to as "high speed sampling data"). To remember).

The sampling frequency is set to 1 MHz in order to detect abnormal discharge such as micro arcing, which occurs and ends in several microseconds. On the other hand, the sampling frequency in the data logger board 55 can be selected within the range of 500 kHz to 5 MHz.

The data logger board 55 also has an internal clock, and the high-speed sampling data is stored in the data logger board 55 as time series data corresponding to the internal clock, and moved and stored in the HDD 61 at regular intervals. . In this way, overflow of the accumulated data in the data logger board 55 can be prevented.

The internal clock of the data logger board 55 and the internal clock of the trigger generation time counter 54 are synchronized. The high-speed sampling data exhibits characteristics due to the abnormality in the time when the abnormality actually occurs in the plasma processing apparatus 2, but the time between the time when the abnormality actually occurs in the plasma processing apparatus 2 and the trigger occurrence time is described above. As time lag occurs. In the PC 50, as described later, high-speed sampling data is handled in consideration of this time lug.

The monitor signal is converted into digital data (hereinafter referred to as "low speed sampling data") with a sampling frequency of the data logger board 56 as 10 kHz, for example.

The data logger board 56 also has an internal clock synchronized with the internal clock of the trigger generation time counter 54. The low-speed sampling data is stored in the data logger board 56 as time series data corresponding to the internal clock. The data is stored in the HDD 61 at regular intervals. In this way, overflow of accumulated data in the data logger board 56 can be prevented.

Thus, in the abnormality detection system 100, detailed data for specifying the abnormality which occurred in the plasma processing apparatus 2 is acquired as a high speed sampling data. However, the data amount of the high speed sampling data is enormous, and analyzing all the high speed sampling data requires a lot of processing cost and time. In addition, in consideration of the time lug between the time at which the abnormality actually occurred and the trigger generation time, the plasma processing apparatus 2 also needs to reliably take out the characteristic showing the abnormality indicated in the high-speed sampling data.

Therefore, in the PC 50, roughly, when the trigger generation time counter 54 receives the representative trigger signal and determines the trigger generation time, a predetermined period before and after the trigger generation time among the high-speed sampling data stored in the HDD 61 is determined. The data of the (time width) is cut out for each of the four sensor signals (hereinafter, the data cut out in this manner is referred to as "range-limited data") and analyzed. As a result, it is judged whether or not an abnormality actually occurred, and what abnormality specifically occurred when an abnormality occurred. The PC 50 executes the abnormality detection / analysis program for performing such a series of processes.

In this way, by using the limited data for the abnormality detection / analysis processing, it is possible to efficiently perform high-accuracy abnormality detection while reducing the processing cost and reducing the processing cost, thereby enabling real-time abnormality detection / analysis processing. . In addition, since the determination of the trigger occurrence time, the collection of the high-speed sampling data, and the abnormality detection / analysis processing are executed in different threads, the delay of each processing can be prevented. Details of the abnormality detection / analysis processing method in the PC 50 will be described later.

When the abnormality detection system 100 determines that an abnormality that is considered to be an obstacle to the operation of the plasma processing apparatus 2 occurs by analyzing the high-speed sampling data, the abnormality detection system 100 causes an alarm to be issued to the plasma processing apparatus 2. The control signal for postponing the start of processing of the next wafer W can be transmitted to the plasma processing apparatus 2.

The PC 50 is connected to a knowledge DB (Knowledge DataBase) 63 in which various data used for the abnormality detection / analysis processing and the related information of the abnormality detection / analysis processing result and furthermore the related information of the abnormality detection / analysis processing result are stored. have.

Among the high-speed sampling data stored in the HDD 61, data other than the range-limiting data is basically unnecessary. Therefore, after the range-limiting data is determined, the data is deleted from the HDD 61, and the range-limiting data is appropriately determined by the HDD ( 61, it is moved to knowledge DB 63 and preserved.

In the knowledge DB 63, the result of the abnormality detection / analysis process and the process conditions (recipe data) used for the processing of the wafer W are linked to and stored in the limited data. In the knowledge DB 63, the type, cause, countermeasure, etc. of the abnormality generated in the plasma processing apparatus 2 can be stored by linking the result of the abnormality detection / analysis process.

The various data stored in the knowledge DB 63 are used to set various parameters (for example, the definition of a normal model, a model for abnormal pattern authentication, a definition of various threshold values, etc.) to be used for subsequent abnormality detection / analysis processing. It helps.

On the other hand, it is also preferable that the abnormality detection system which is the same form as the plasma processing apparatus 2 and accompanies the plasma processing apparatus arrange | positioned elsewhere can be made into the structure which can access the knowledge DB 63 using a communication line. As a result, information about an abnormality occurring in the plasma processing apparatus disposed elsewhere can be accumulated in the knowledge DB 63 to assist in the abnormality detection / analysis processing of the plasma processing apparatus 2. In addition, the abnormality detection system accompanying the plasma processing apparatus arrange | positioned elsewhere can easily cope with the abnormality which generate | occur | produced in the plasma processing apparatus by extracting necessary information from the knowledge DB 63. FIG.

Next, the abnormality detection / analysis processing method by the abnormality detection system 100 is demonstrated. First, the outline of the abnormality detection / analysis processing method will be described, and then, the key processing among the abnormality detection / analysis processing will be described in detail.

4 is a flowchart showing an outline of an operation of the abnormality detection system 100.

At the same time as the start of the new plasma processing device 2 or immediately after maintenance, the abnormality detection system 100 starts. At this time, since the apparatus condition of the plasma processing apparatus 2 may be unclear or changed, the ultrasonic sensor 41 and the high frequency power sources 19 and 21 are first performed by a pilot run (test production). Noise level recognition by the monitor sensor (step S1).

If the noise level measured in step S1 is compared with the normal model and is within the allowable range, the S / N ratio and the like of the normal model are tuned based on the measurement result of the pilot run to define the normal model (step S2). On the other hand, the normal model is defined in time series data from each ultrasonic sensor 41 in a defined section such that the plasma processing apparatus 2 is in an idle state, during plasma generation, or other (for example, during wafer transfer). It is a waveform defined by the statistical value after performing a filter process etc., for example, parameters, such as a maximum, minimum, average, and variance.

On the other hand, although not shown in Fig. 4, when the noise level measured in step S1 shows any device abnormality or the difference with the previous normal model is large, an alarm is triggered and the plasma processing apparatus 2 is locked. This can be checked by the operator or administrator.

After the normal model is defined in step S2, tuning of the threshold for waveform recognition is performed (step S3). For example, the threshold value for generating a trigger signal in the trigger 52, and the threshold value for peak determination and feature amount extraction in a waveform represented by down-sampling data described later are determined.

Subsequently, the abnormal pattern recognition model stored in the HDD 61 (or the knowledge DB 63) is uploaded (step S4). The abnormal pattern recognition model is various feature quantities representing waveforms, and is used for pattern recognition with waveforms based on range-limited data (high-speed sampling data) in the PC 50 and stored in, for example, the RAM 58.

On the other hand, in step S4, one or a plurality of abnormal pattern recognition models can be uploaded for each abnormality whose cause is found, and also an abnormal pattern recognition model which shows a waveform whose cause is unknown as an abnormality detected in the past is uploaded. can do.

Next, the recognition rate is confirmed and evaluated by simulating the abnormal pattern recognition model (step S5). Step S5 can be executed using the high speed sampling data and the low speed sampling data obtained in the pilot run, but may be executed by newly performing a pilot run. When a constant recognition rate (for example, 90%) is obtained in step S5, the processing for the new start and the start immediately after maintenance is terminated, so that normal operation of the plasma processing apparatus 2 and the abnormality detection system 100 can be performed. (Step S6).

On the other hand, although not specified in Fig. 4, if a constant recognition rate is not obtained in step S5, steps S2 to S4 are repeatedly performed until a constant recognition rate can be obtained.

Subsequently, when the plasma processing apparatus 2 is operated, the abnormality detection system 100 monitors the plasma processing apparatus 2 under the conditions set up to step S5 (processing starts at the time of operation continuation). If the abnormality detection system 100 detects an abnormality during the operation of the plasma processing apparatus 2 (if the trigger 52 generates a trigger signal), the abnormality detection / analysis process for specifying the abnormality is executed (step). S7).

With respect to the result obtained in the abnormality detection / analysis process of step S7, it is judged whether abnormality determination is made correctly (step S8). For example, when the recognition rate of pattern recognition with the abnormal pattern recognition model is low, it is set as "the problem is abnormal in determination" (NO at step S8), and reset of various parameters is performed (step S10).

In step S10, for example, tuning of a threshold value, addition of a waveform feature amount, addition of a model for abnormal pattern recognition, and the like are performed. Subsequently, redefinition of the normal model, confirmation and evaluation of the recognition rate are performed (step S11). In step S11, when a constant recognition rate is obtained, the process returns to step S7 (abnormal detection / analysis processing).

Although not shown in Fig. 4, if a constant recognition rate is not obtained in step S11, the process is repeated until steps S10 to S11 can obtain a constant recognition rate.

If it is determined in step S8 that there is no problem in the abnormality determination ("YES" in step S8), the abnormality detection system 100 continues to operate (step S9), so that the monitoring of the plasma processing apparatus 2 is performed by plasma. It is made until the operation stop of the processing apparatus 2 (end of processing at the time of continuous operation).

Next, the detailed procedure of step S7 (abnormal detection / analysis processing) will be described.

5 is a flowchart showing a schematic procedure of an abnormality detection / analysis processing method.

Initially, information acquisition of the process conditions (recipe data) of the wafer W is performed (step S21). For example, when it is desired to detect only an abnormality occurring during plasma generation, by obtaining the process conditions, it is possible to acquire the high speed sampling data and the low speed sampling data only for the period from the start of generation of plasma to the end.

When reception of the sensor signal from the ultrasonic sensor 41 is started, the sensor signal is distributed by the splitter 65 into two identical signals, one of which is triggered after the frequency band is narrowed by the filter 51a. ) Is input to the PC 50 (data logger board 55) after the frequency band is narrowed by the filter 51b (step S22).

When the trigger 52 detects that the received sensor signal has a peak having a maximum value equal to or greater than a predetermined threshold, the trigger 52 generates a trigger signal and outputs the trigger signal to the OR circuit 53. Upon receiving the trigger signal, the OR circuit 53 generates a representative trigger signal in accordance with the second trigger signal processing method described above and outputs the representative trigger signal to the PC 50 (step S23).

The sensor signal directly input to the PC 50 after step S22 is converted into high-speed sampling data by the data logger board 55, temporarily stored, and transferred and stored in the HDD 61 at regular intervals. When the trigger generation time counter 54 receives the representative trigger signal from the OR circuit 53, the trigger generation time is determined, and the limited data is cut out from the high speed sampling data (step S24), and the unnecessary data other than the limited data. Is deleted.

In the period of the range-limited data cut out in step S24, considering that the AE caused by the abnormality generated in the plasma processing apparatus 2 is attenuated for a certain period after generation, and that the trigger signal is generated quickly after generation, For example, it is preferable to set a short time before the time of the trigger generation time in consideration of the integration period, and set a long time so that the peak waveform does not break in the middle of the time after the trigger time.

Subsequently, waveform analysis of the limited data is performed (step S25). Details of the waveform analysis method are described later. Since the process of step S25 needs to be performed on the range-limited data cut out for all the representative trigger signals generated by the OR circuit 53, it is determined whether or not the number of received representative trigger signals has been processed (step S26).

When the processing by the number of received representative trigger signals is completed (YES in step S26), abnormal pattern determination is performed (step S27). At this time, even if abnormality is difficult to be identified only by the sensor signals of one ultrasonic sensor 41, by comparing the respective sensor signals obtained from the four ultrasonic sensors 41, the probability of identifying the cause of the abnormality can be greatly increased. .

On the other hand, the processes of steps S24 and S25 are repeated until the processing for the number of received representative trigger signals ends (NO in step S26).

Based on the determination result of step S27, it is determined whether an abnormality actually occurred (step S28). When the occurrence of the abnormality is confirmed ("YES" in step S28), the PC 50 transmits a signal to the plasma processing apparatus 2 to execute an alarm alarm or the next process stop (step S29). The detection / analysis process ends.

On the other hand, when an abnormal occurrence is not confirmed and when an abnormal occurrence is acknowledged but it is determined that no action such as alarm alarm or next process stop is necessary (NO in step S28), the abnormality detection / analysis process is terminated. On the other hand, at the end of the abnormality detection / analysis process, various data obtained in steps S24, S25, and S27 are stored in the knowledge DB 63.

The above-described steps S23, S24, S25 and S27 will be described in more detail below.

6 is a flowchart showing the detailed procedure of step S23 (trigger output).

The sensor signal passing through the filter 51a is input to the trigger 52, and a simple preprocess (here, digital sampling process at 10 kHz) is performed (step S31).

Based on the sampling data obtained in step S31, it is determined whether or not there is a peak in the sensor signal at which the maximum value is greater than or equal to the threshold (step S32). If there is a peak whose maximum value is greater than or equal to the threshold (" YES " in step S32), in order to further distinguish the peak from a peak such as noise, whether the peak continues for a certain period of time, i.e., a time width of a certain period or more in the peak It is judged whether or not there is (step S33).

If the peak continues for a certain period or more (" YES " in step S33), the trigger 52 generates a trigger signal and outputs it to the OR circuit 53 (step S34). In the OR circuit 53, it is determined whether the integration period has elapsed from the first trigger signal (step S35), and waits for the completion of the integration period ("NO" in step S35).

When the integration period has elapsed in step S35 ("YES" in step S35), the OR circuit 53 combines the trigger signals received within the integration period to generate a representative trigger signal (step S36), so that the PC 50 Output (step S37).

After outputting the representative trigger signal in step S37, the first trigger signal received becomes a reference trigger signal for generating the next representative trigger signal. If the judgment of steps S32 and S33 is "NO", monitoring continues (step S38), and the process of step S23 ends by the completion of steps S37 and S38.

7 is a flowchart showing the detailed procedure of step S24 (waveform cut / preserve).

When the trigger generation time counter 54 equipped with the PC 50 receives the representative trigger signal (step S41), it determines the trigger generation time (step S42). When the trigger generation time is determined in step S42, by applying a predetermined period to the trigger generation time, the data surrounding the trigger generation time for the high speed sampling data and the low speed sampling data (creation of the range-limited data) is performed ( Step S43). On the other hand, the "predetermined period" is determined in consideration of the integration period required for generation of the representative trigger signal.

The range limiting data cut out in step S43 is stored in the HDD 61 (step S44), and the unnecessary high speed sampling data and the low speed sampling data are erased from the HDD 61.

Subsequently, analysis of the limited data (waveform analysis: step S25) is performed, the time series data obtained in the step S25 is stored in the HDD 61 (step S45), and appropriately transferred to the knowledge DB 63, Are preserved. In addition, the HDD 61 stores the analysis result (the characteristic amount of the peak indicating abnormality) obtained in step S25 (step S46), and the waveform cutting / saving process is completed.

8 is a flowchart showing the detailed procedure of step S25 (waveform analysis).

There are four range limitation data for one representative trigger signal. Therefore, for each of the range-limited data, it is judged whether or not the peak value (maximum amplitude) is equal to or greater than the predetermined threshold value (whether there is a peak having a peak value equal to or greater than the threshold value) (step S51). If the peak value is less than the threshold ("NO" in step S51), it is determined that it is not a meaningful waveform (step S55), and the waveform analysis process regarding the range limitation data is completed.

If the peak value is equal to or greater than the threshold value (YES in step S51), representative value extraction (down sampling) is performed (step S52). Since the sensor signal generated by the ultrasonic sensor 41 exhibits a vibration waveform in which the voltage value changes between a positive value and a negative value, for example, the maximum amplitude (absolute value) can be employed as the representative value. In this case, in step S52, down sampling is performed by converting the negative value into a positive value, superimposing the original plus value, and connecting the maximum amplitude value of the waveform thus obtained with a sampling frequency of 10 kHz, for example. Process into data.

On the other hand, the representative values that can be adopted in step S52 include, in addition to the maximum amplitude, the minimum amplitude, the average amplitude, and the like, and should be selected according to the characteristics of the detection target signal. By reducing the number of data in step S52, the data processing load in subsequent steps S53 to 58 can be reduced, and the data processing time can be shortened.

Next, based on the obtained waveform of the down sampling data, it is judged whether or not the peak continues for a predetermined period or more (step S53). For example, it is determined whether the height of the waveform is maintained at a height of 15% or more of the maximum amplitude over a period of time or more.

If the peak does not continue for a certain period or longer ("NO" in step S53), it is determined that the waveform is not meaningful (step S55), and the waveform analysis process for the down-sampling data and the range-limited data that is causing it is performed. Quit.

If the peak continues for a certain period or more ("YES" in step S53), it is determined that there is a meaningful waveform (step S54), and the time (time for specifying the waveform) of the waveform is estimated (step S56). . As this time estimation method, energy monitoring, cross-correlation value monitoring, local normal AR model, etc. are mentioned in order of data processing time being short. When the energy monitoring method is used, the time showing the maximum amplitude can be the time of the waveform.

Thereafter, the waveform feature amount is extracted from the down sampling data (step S57). As the waveform characteristic quantity, the maximum energy (time integration value of the maximum amplitude), the time of showing the maximum energy, and the time of reaching the maximum energy (the first time of counting forward of the time of showing the maximum energy, for example, below 25% of the maximum energy) , The maximum energy dissipation time (counting backwards from the time at which the maximum energy is shown, the first time below the 25% of the maximum energy), the intermittent / continuous wave (less than 25% of the maximum energy for a given period of time) Continuous waves when there is no work, and intermittent waves when there is a step).

The analysis target (range) is narrowed down from the estimated time of the waveform obtained in step S56 and the waveform feature quantity obtained in step S57 (step S58). Then, the waveform feature amount is extracted from the limited data for the analysis target narrowed in step S58 (step S59). Specifically, the waveform characteristics are revealed by performing fast Fourier transform (FFT) on the range-limited data (sampling frequency: 1 MHz).

As the waveform characteristic quantity, the FFT start time (= maximum energy arrival time), the FFT end time (the first time when the FFT samples are multiplied by 2 from the maximum energy dissipation time) and the FFT sample Number (the number of samples used in the FFT, for example, 16348 can be the upper limit), maximum peak frequency (frequency with maximum amplitude), average frequency (frequency where peak area (energy) exceeds 50% of total peak area) And the ratio of the reference frequency or more (the ratio of the reference frequency (for example, 20% of the sampling frequency) or more). In this way, the waveform analysis process is completed.

9 is a flowchart showing the detailed procedure of step S27 (abnormal pattern determination).

First, the process processing condition is read out (step S61). Since the abnormality that can occur based on the read process processing conditions can be limited, the model for abnormal pattern recognition can be compressed, and the determination time can be shortened. Moreover, the accuracy at the time of determining the abnormality which generate | occur | produced can be improved.

Subsequently, the waveform feature amount obtained in the waveform analysis (step S25) is read (step S62), and the abnormal pattern recognition model is read (step S63). The pattern recognition algorithm which compares the waveform feature amount read in step S62 with the abnormal pattern recognition model read in step S63 determines whether any abnormality has occurred (step S64). As a pattern recognition algorithm, a well-known method, for example, a support vector machine (SVM) etc. can be extended and used for multistage determination. In this way, the abnormal pattern determination processing is finished.

Next, the modification of the abnormality detection system 100 is demonstrated.

<Modification 1>

The configuration of the trigger generation time counter 54 and the data logger boards 55 and 56 is not limited to the above embodiment, and the clock control of the trigger generation time counter 54 and the data logger boards 55 and 56 is 1. It may be performed by two external reference synchronization clocks.

<Modification 2>

In the above embodiment, the OR circuit 53 is configured to integrate a plurality of trigger signals into one representative trigger signal (second trigger signal processing method), but the following method is used as a method of generating the representative trigger signal. It is also preferable.

In other words, instead of the trigger occurrence time counter 54, the PC 50 is provided with a buffer port and a time counter so that the external reference synchronization clock gives time information to the time counter. The OR circuit 53 sequentially outputs the trigger signal received from the trigger 52 to the buffer port in time series according to the first trigger signal processing method described above. In parallel with this, the counter value output from the time counter is taken into the buffer port. In this way, a counter value is provided to the trigger signal.

The buffer port integrates a plurality of trigger signals within a certain period into one representative trigger signal. At this time, since each trigger signal has time information, as the trigger generation time of the representative trigger signal, for example, time information of the first trigger signal within a predetermined period can be used.

Therefore, when the above-described second trigger signal processing method is used, the trigger generation time of the representative trigger signal is that the integration period has elapsed from the generation time of the first trigger signal that generates the representative trigger signal. The trigger generation time of a representative trigger signal can be made into the time which became close to the time which the abnormality actually occurred in the plasma processing apparatus 2.

<Modification 3>

As another generation method of the representative trigger signal in the OR circuit 53, the trigger signal received at the most is output to the PC 50 as the representative trigger signal, and thereafter, until a time corresponding to the integration period elapses. In addition, a method of not outputting the received trigger signal to the PC 50 can be used. Also in this method, the trigger occurrence time determined by the trigger occurrence time counter 54 can be made into the time which became close to the time which the abnormality actually occurred in the plasma processing apparatus 2.

<Modification 4>

In the above embodiment, the cutting of the limited data is performed from the high speed sampling data stored in the HDD 61. However, the present invention is not limited to this, and the high speed sampling data is stored in the data logger board 55. Range-limited data is cut in accordance with the reception of the signal, and data other than the range-limited data is erased from the data logger board 55 while moving and storing the cut-out range-limited data from the data logger board 55 to the HDD 61. A configuration may be made. The same method can be used for the low-speed sampling data.

In addition, when the limited data is specified from the high-speed sampling data and the unnecessary data is deleted, the deletion may be performed after confirming that there are no peaks above a certain threshold in the unnecessary data. Further, even if the limited data is specified, the unnecessary data may not be erased for a predetermined time, or the high-speed sampling data may be periodically transferred to the HDD 61 to be stored for a certain period of time, and the unnecessary data may be appropriately deleted.

<Modification 5>

In the above-described embodiment, in step S51 of the waveform analysis process, the range-limited data determined that the peak value is less than the predetermined threshold value is excluded from subsequent analysis objects, and in step S53, the peak does not continue for a certain period or more. Although the downsampling data and the range-limiting data that cause it are excluded from subsequent analysis, the four range-limiting data corresponding to one representative trigger signal are regarded as a group, and the group has a threshold value over which a peak value is defined. In the case where there is range-limited data determined to be, and when there is any down sampling data whose peak continues for a certain period of time, the processing method is performed to proceed to the next step for all data in the group. It may be configured. Thereby, the mutual relationship of the sensor signals from the four ultrasonic sensors 41 can be grasped.

<Modification 6>

In the plasma processing apparatus 2, abnormal discharges generated during plasma generation also have a high probability of appearing in the monitor signal. Therefore, the structure which generates a trigger signal based on a monitor signal may be sufficient. Further, the same process as that for the sensor signal may be performed on the monitor signal, and the cause of the abnormality may be determined and determined from not only the sensor signal but also the characteristic amount of the waveform indicated by the monitor signal.

<Modification 7>

When high-speed sampling the sensor signal at 1 MHz, the thinning data which is substantially the same as the sampling at 10 kHz is created simultaneously with the high-speed sampling data, and the data until the analysis target is narrowed down by time estimation (step S58). The processing may be performed using this sound data.

However, it is preferable to limit this method only to the case where the peak value determination of step S51 using 1 MHz sampling data is empirically confirmed, even if 10 kHz sampling data is performed.

As mentioned above, although embodiment of this invention was described, this invention is not limited to the form mentioned above. An object of the present invention is to supply a storage medium storing program codes of software for realizing the functions of the above embodiments to a PC 50 or an external server, and the CPU of the PC 50 or an external server stores the storage medium. It is also achieved by reading and executing the program code stored in the program code.

In this case, the program code itself read out from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, a hard disk, a magneto-optical disk, a CD-ROM, a CD-R, a CD-RW, a DVD-ROM, a DVD-RAM, a DVD- RW, DVD + RW, magnetic tape, nonvolatile memory card, ROM, etc. can be used. Alternatively, the program code may be downloaded via a network. In this case, the program code is supplied by downloading from another computer or database (not shown) connected to the Internet, a commercial network or a local area network or the like.

In addition, by executing the program code read by the CPU, not only the function of the above embodiment is realized, but also the OS or the like running on the CPU performs part or all of the actual processing based on the instruction of the program code. It also includes a case where the functions of the above-described embodiments are realized by the processing.

Furthermore, the program code read out from the storage medium is recorded in the memory provided in the function expansion board inserted into the PC 50 or the external server or the function expansion unit connected to the PC 50 or the external server, and then the program code. Based on the instruction of the above, the function expansion board, the CPU provided in the function expansion unit, etc. perform part or all of the actual processing, and the processing includes the case where the functions of the above-described embodiments are realized.

The program code may be in the form of an object code, a program code executed by an interpreter, script data supplied to an OS, or the like.

In the plasma processing apparatus according to the above-described embodiment, the monitor signals from the high frequency power sources 19 and 21 are used in addition to the sensor signals from the ultrasonic sensors in order to detect the plasma abnormal discharge. Without using this monitor signal, a separate signal, for example, a current value monitor for measuring the value of the current flowing through the susceptor or the electrode plate for absorbing the wafer W, and a reflected wave monitor for measuring the reflected wave of high frequency power from the susceptor And a monitor signal from a phase monitor for measuring the phase variation of the high frequency power.

In the above embodiment, the case where the abnormality detection system is applied to an etching apparatus which is a kind of plasma processing apparatus has been described. However, the abnormality detection system can be applied to other plasma processing apparatuses such as a CVD film deposition apparatus and an ashing apparatus. The present invention is not limited to the plasma processing apparatus but may be applied to a coating and developing apparatus, a substrate cleaning apparatus, a heat treatment apparatus, an etching apparatus, and the like.

In the above-mentioned embodiment, although the wafer W was mentioned as a board | substrate to be processed, the to-be-processed substrate is not limited to this, A glass substrate, such as a flat panel display (FPD), may be sufficient.

Claims (11)

As an abnormality detection system which detects the abnormality which arises in a processing apparatus,
A plurality of ultrasonic sensors for detecting acoustic emission generated by the processing apparatus;
A distribution unit for distributing each output signal of the plurality of ultrasonic sensors into a first signal and a second signal, respectively;
A trigger generation unit for sampling the first signal at a first frequency and generating a trigger signal when a predetermined characteristic is detected;
A trigger generation time determination unit which receives the trigger signal to determine a trigger generation time;
A data creation unit which creates sampling data obtained by sampling the second signal at a second frequency higher than the first frequency;
And a data processing unit for analyzing abnormality occurring in the processing apparatus by performing waveform analysis of data corresponding to a predetermined period of time based on the trigger generation time determined by the trigger generation time determination unit among the sampling data. Abnormality detection system. A trigger signal processing unit according to claim 1, further comprising a trigger signal processing unit for integrating the plurality of trigger signals into a single signal as a representative trigger signal when a plurality of the trigger signals are generated within a predetermined period of time,
And the trigger generation time determination unit determines the trigger generation time with respect to the representative trigger signal. The abnormality detection system according to claim 1 or 2, further comprising a filter for removing noise from each output signal of the plurality of ultrasonic sensors. The method according to any one of claims 1 to 3, wherein the first frequency is 10 kHz-5 MHz,
And the second frequency is 500 kHz to 5 MHz. As an abnormality detection method which detects the abnormality which arises in a processing apparatus,
A detection step of detecting acoustic emission generated by the processing apparatus by a plurality of ultrasonic sensors;
A distributing step of distributing each output signal from the plurality of ultrasonic sensors obtained in the detecting step into a first signal and a second signal by a distributing unit, respectively;
A trigger signal generation step of sampling the first signal by the A / D conversion unit at a first frequency and generating a trigger signal by the signal generation unit when a predetermined feature is detected;
A trigger generation time determining step of receiving the trigger signal and determining a trigger occurrence time of the trigger signal by a time counter unit;
A sampling data creation step of sampling the second signal at a second frequency higher than the first frequency by an A / D conversion unit to create sampling data;
And a data processing step of analyzing an abnormality occurring in the processing device by performing a waveform analysis of data corresponding to a predetermined period of time based on the trigger generation time determined in the trigger generation time determination step of the sampling data. An abnormality detection method characterized by the above-mentioned. The method according to claim 5, further comprising a trigger signal processing step of integrating the plurality of trigger signals into one signal as a representative trigger signal when a plurality of the trigger signals are generated within a predetermined period of the trigger signal generation step. ,
In the trigger generation time determining step, the trigger generation time is determined with respect to the representative trigger signal. 7. The abnormality detecting method according to claim 5 or 6, further comprising a noise removing step of removing noise by a filter from the first signal and the second signal obtained by the distributing step. The method according to any one of claims 5 to 7, wherein the first frequency is 10 kHz to 5 MHz,
An abnormality detection method characterized in that the second frequency is 500 kHz to 5 MHz. The method of claim 5, wherein the data processing step comprises:
A cutting step of cutting out data corresponding to the predetermined period from the sampling data;
A first waveform feature variable extracting step of extracting a waveform feature quantity from the down sampling data when a meaningful waveform exists in the generated down sampling data by performing down sampling by a representative value on the data cut out in the cutting step; Wow,
By estimating the time of the waveform feature variable extracted by the first waveform feature variable extracting step, the analysis target of the data cut out in the cutting step is narrowed, and the waveform feature variable is extracted from the data cut out in the cutting step with respect to the analysis target. Extracting a second waveform feature variable;
And a determination step of determining an abnormality occurring in the processing apparatus by performing pattern recognition between the waveform characteristic quantity obtained in the second waveform feature quantity extracting step and a preset abnormal pattern recognition model. The method according to any one of claims 5 to 9, further comprising a process condition acquiring step of acquiring a process condition of a predetermined process executed in the processing apparatus, which is performed before the detecting step,
And the detecting step is executed only during an execution period of the predetermined process included in the process condition acquired in the process condition obtaining step. A computer-readable storage medium storing a program for executing an abnormality detecting method for detecting an abnormality occurring in a processing apparatus determined by a computer-controlled abnormality detecting system, comprising:
The abnormality detection method,
A detection step of detecting acoustic emission generated by the processing apparatus by a plurality of ultrasonic sensors;
A distribution step of distributing each detection signal from the plurality of ultrasonic sensors obtained in the detection step into a first signal and a second signal, respectively, by a distribution unit;
A trigger signal generation step of sampling the first signal by the A / D conversion unit at a first frequency and generating a trigger signal by the signal generation unit when a predetermined feature is detected;
A trigger generation time determining step of receiving the trigger signal and determining a trigger occurrence time of the trigger signal by a time counter unit;
A sampling data creation step of sampling the second signal at a second frequency higher than the first frequency by an A / D conversion unit to create sampling data;
A data processing step of analyzing an abnormality occurring in the processing apparatus by the computer performing waveform analysis of data corresponding to a predetermined period of time based on the trigger generation time determined in the trigger generation time determination step of the sampling data; A computer-readable storage medium storing a program for executing the abnormality detecting method.

Images